Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

They probably did not batch. It’s realistic they will have issues if code is written to handle 1 job at a time and needs to make several roundtrips to the same db inside the same locking transaction.

Leases exist for a reason.





> if code is written to handle 1 job at a time and needs to make several roundtrips to the same db inside the same locking transaction.

Do you mean the application code? The worker itself causing the bottleneck is definitely one possibility however if that were the case the issue wouldn't have resolved itself when they switched to a different job queue.


Well, you would no longer have thousands of open transactions maintaining the locks.

Of course, if q on top of psql is reasonably implemented (lease), what they say makes no sense.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: